OpenAI begins a network for red teaming to strengthen its models

OpenAI has launched the OpenAI Red Teaming Network, a group of contracted experts who will assist the company in assessing and mitigating risks associated with its AI models. Red teaming, which can identify biases in AI models, is becoming an increasingly essential part of AI model development. Although OpenAI has previously involved external experts in testing its models, the Red Teaming Network will formalize these efforts. Members of the network will provide their expertise at various stages of model and product development and will also have the chance to discuss general red teaming practices and findings amongst themselves. OpenAI is calling for experts from a wide range of fields to join the network, stressing that they value different perspectives on the assessment of AI system impacts.